I was contacted to make an ardupilot drone based on a Ai rendering I believe from midjourney the image was good looking and had sifi look but needed to be simple in order to make it’s reality clone. I think this could be the first ai to 3d to real craft made this week for Ardupilot.
But why? This pseudo-AI has brain of a parrot and basicly just approximate best match from large database, knowing s**t about aerodynamics, design principles, requirements and stuff
2 Likes
I was kind of wondering the same thing. MidJourney is fine for creating an aesthetic, if that’s what you’re going for.
Incorporating AI generative design where you provide physical constraints and requirements would be a far more interesting exercise than just trying to match a pretty picture.
1 Like
I think the best part is to have something new to absorb pushing imagination. In my case if a client asks me to do it I have to try.
I completely understand complying with a client’s needs/desires, but I hesitate to call this innovative use of AI.
1 Like
1000 points of lights. lol
Reality of Quads they don’t really need to be super aerodynamic at the speed they travel I was told.
1 Like
My background is TV for Star Trek so I find ai important futuristic tool.
I agree that AI can be an outstanding tool. For example, this GoPro mount was created using Fusion360’s generative design tool, providing constraints for mounting points and positioning as well as optimizing for weight and strength:
I disagree that MidJourney is of much utility for air vehicle design. It seems a horrific misapplication, much like asking ChatGPT to write unique/publishable content. There are applications for both, but many folks have no earthly idea how to leverage the capabilities well.
1 Like
Why is this a bad use of ai?
“I disagree that MidJourney is of much utility for air vehicle design. It seems a horrific misapplication,”
Why do you think ai is a bad tool.
I never said that AI is a bad tool. I am asserting that people rampantly misuse it and gave an example where I think I leveraged it well.
Let me pitch that right back at you - why this an example of it being a good tool? The camera position looks useless, likely with much of its FOV obscured by the frame and props. The camera housing doesn’t appear large enough to house any other electronics, forcing the battery and autopilot into the ungainly pod that appears to be in a suboptimal position for CG. So, the client created a picture he/she deemed “neat-o” and asked you to make it real…but it appears there is zero consideration given to actual utility or performance, nor could there be, since MidJourney literally just makes “pretty” pictures.
1 Like
"The camera position looks useless, likely with much of its FOV obscured by the frame and props. The camera housing doesn’t appear large enough to house any other electronics, "
turns out it’s in a good position.
“forcing the battery and autopilot into the ungainly pod that appears to be in a suboptimal position for CG”
The battery is placed vertically a improved CG location!
POV from the GoPro:
Those are human optimizations of a rendering by an AI that could not possibly have understood or accounted for any of it. Again, misapplication just to create some bizarre aesthetic. We can agree to disagree. I’m curious what others think.
You did an outstanding job bringing that image to life, so credit where it’s due.
Don’t mistake the position I’m taking. My day job has me deeply involved in the use of AI within the industry space where I work. I’m speaking from a position of understanding and innovation, not ignorance or closed-mindedness.
1 Like
My last job was in ai for rail road safety. “understanding and innovation, not ignorance or closed-mindedness.” then why the blow back?
I feel like I’ve supported my argument pretty well here, but let me try and break it down further:
MidJourney is an AI “art” creator. It just makes pictures (often of people with 6 fingers…). It has no knowledge of the physical world and just regurgitates pixels in a way it thinks we will find pleasing (or displeasing, if so prompted). Again, it’s just a thing to create aesthetics, and aesthetics don’t fly. It can’t optimize performance, space, weight, lift, thrust, drag, or…anything. Sure, it can give us visual inspiration, but I fail to see how that’s even innovative - we could just as easily ask a child to draw a cool picture and draw inspiration from that. I stand by my assertion that MidJourney is misapplied as an aircraft design tool.
On the other hand, you haven’t really justified what’s so good about it. By outward appearance, your client wants to make a Fisher Price drone toy for kids and found a pretty picture they think they can market. That doesn’t exactly make for a solid use case of MidJourney in a technical space.
1 Like
Let me respond. Most fames i build work very well this is no exception. If the rig flew badly I would of tabled the design. Yes I make toys and some are published world wide.
Hmm. The client severed in the NAVY top in his class and also builds drones no need to make assumptions. I actually have been thinking that Ardupilot is no longer a safe place for publicly sharing ideas. not directed at you you seem overall nice. But the ratio to kindness is not what I signed up for.
One thing ai visually searching 1000’s of images including jets like the Blackbird. can’t really be a bad thing.
If it seemed like I was beating you up personally, I do truly apologize. It’s a fascinating conversation, and one I’m obviously a little passionate about. Always cool to see new ideas…even if I disagree with the premise 
And I’ve been known to make plenty of well founded assumptions about those Navy guys…
No worries!